Consolidating Inner Product Masking

Masking schemes are a prominent countermeasure to defeat power analysis attacks. One of their core ingredients is the encoding function. Due to its simplicity and comparably low complexity overheads, many masking schemes are based on a Boolean encoding. Yet, several recent works have proposed masking schemes that are based on alternative encoding functions. One such example is the inner product masking scheme that has been brought towards practice by recent research. In this work, we improve the practicality of the inner product masking scheme on multiple frontiers. On the conceptual level, we propose new algorithms that are significantly more efficient and have reduced randomness requirements, but remain secure in the t-probing model of Ishai, Sahai and Wagner (CRYPTO’03). On the practical level, we provide new implementation results. By exploiting several engineering tricks and combining them with our more efficient algorithms, we are able to reduce execution time by nearly 60% compared to earlier works. We complete our study by providing novel insights into the strength of the inner product masking using both the information theoretic evaluation framework of Standaert, Malkin and Yung (EUROCRYPT’09) and experimental analyses with an ARM microcontroller.